Sphinx Guideline version 1

3.4 Decoding With Sphinx
As mentioned above, you need the following files for decoding:
Trained models
Dictionary
Filler dictionary
Language model/ FSG
Test data
To get a list of arguments in Sphinx3 decode, just type sphinx3_decode in the command line. You
will get a list of arguments and their usage.
In general, the configuration file needs, at the very least, to contain the following arguments:
-hmm followed by the acoustic model directory
-dict followed by the pronunciation dictionary file
-fdict followed by filler dictionary
-lm followed by the language model, OR
-fsg followed by the grammar file
-ctl followed by the control file (fileids)
-mode fsg (for FSG, for n-gram let it be the default mode, so need to specify this for n-gram)
However, in this case, there are several additional arguments that are necessary:
-adcin yes (tells the recognizer that input is audio, not feature files)
-adchdr 44 (tells the recognizer to skip the 44-byte RIFF header)
-cepext .wav (tells the recognizer input files have the .wav extension)
-cepdir wav (tells the recognizer input files are in the 'wav' directory)
-hyp followed by the output transcription file
In case your training arguments are different from default arguments, then set it manually by
providing the following parameters.
-hypseg \ (followed by Recognition result file, with word segmentations and scores)
-logfn \ (followed by Log file (default stdout/stderr))
-verbose yes \ (Show input filenames)
-lowerf 133.33334 \ (Lower edge of filters)
-upperf 3500 \ (upper edge of filters)
-nfft 256 \ (Size of FFT)
-wlen 0.0256 \ (Hamming window length)
-nfilt 33 \ (Number of filter banks)
-ncep 13 \ (Number of cep coefficients)
-samprate 8000 \ (Sampling rate)
-dither yes (Add 1/2-bit noise)
The parameters like lowerf, upperf, nfft, wlen, nfilt,ncep should be same as used for training for
good results.
23